Runs can currently look passed even when final artifacts are missing or broken.
Current problem:
runPlaytest can mark summary.status = passed after the agent loop.- Stop/recording finalization failures can be swallowed or recorded separately instead of failing the run.
- Remote upload errors are stored as
summary.uploadError, but the process can still exit successfully if the summary status is already passed.
Expected:
A run should fail if the WebM is missing/empty/corrupt, if stop/finalization fails, or if required artifact upload fails. The pass/fail status should describe whether the whole deliverable is usable, not only whether the agent loop returned.
Runs can currently look passed even when final artifacts are missing or broken.
Current problem:
runPlaytestcan marksummary.status = passedafter the agent loop.summary.uploadError, but the process can still exit successfully if the summary status is already passed.Expected:
A run should fail if the WebM is missing/empty/corrupt, if stop/finalization fails, or if required artifact upload fails. The pass/fail status should describe whether the whole deliverable is usable, not only whether the agent loop returned.